I was looking at many players' ELO over time graphs and noticed something interesting. Usually players stay in the same 200-300 bracket through very long periods of time (I'm talking decades here) despite staying active and playing. My intuition would be that experience would have some positive correlation with ELO but it looks like it's not the case. I know many research has been done on this topic but I have never seen this question explored.
To recap: Is there a known positive correlation between experience (games played) and ELO improvement? If there is no such correlation (as I found to be the case through my limited observation), what are the possible explanations?
Add a comment
|
4 Answers
If you are asking why players don't improve despite decades of playing the simple answer is that the more time you spend practicing and reinforcing your mistakes the more ingrained they become and the more likely it becomes that your level will hardly change..
The principle teaching of the Soviet School of chess as described by Dvoretsky and others is that the key to improvement lies in annotating your own games.
Only by critically examining the (bad) moves you make can you identify the kind of mistakes you typically make and only then can you take positive steps to correct them.
This is a very active thing to do. It involves putting your brain into gear. Just putting your game into an engine and seeing where the evaluation changes by a big margin and noting the engine suggestion doesn't count. That is just passive.
EDIT: Analyzing and annotating your own games to identify mistakes with the idea of avoiding those mistakes in the future, something the Soviets practiced in the Botvinnik era, is also one of the centerpieces of Axel Smith's award-winning 2013 book, "Pump up your Rating".
Here is his summary of this technique -
1) Analyse your games with a friend; avoid using an analysis engine
2) Annotate the game in ChessBase, and highlight your mistakes
3) Check the analysis with an engine, and update the annotations
4) Write down the mistakes in a text file
5) Collect all the mistakes in a List of Mistakes, where they are sorted into various categories
6) Make conclusions about typical mistakes, and how they should be avoided in the future
Axel Smith also gives examples of some of his own mistakes and how he used this technique to identify them and correct them in his play.
answered Dec 30, 2017 at 12:33
-
I like your answer a lot. Do you know about statistical studies on the subject? I'm interested in both professional and amateur players data. If there are no such studies is there a good resource for ELO data with some good API ? (i was just looking manually through graphs) Commented Dec 30, 2017 at 13:58
-
I don't think anybody has done analysis on this. Your best source would be the current and historic data that FIDE makes available here - ratings.fide.com/download.phtml. Two thirds of the way down the page there is an option to download archive data going back to Jan 2001. You could download this data, load into your own relational database and then run queries to extract the statistical data you need. This sounds like a lot of work which is probably why nobody has done this yet. Note there is more data going back to 1967 at olimpbase.org.– Brian Towers ♦Commented Dec 30, 2017 at 17:17
-
Using an engine only for blunder-checking does seem very passive, but how about re-analysing one's games by using engines (trying out variations and exploring other lines, etc)? Commented Jan 1, 2018 at 10:39
-
@HarryWeasley I've edited my answer to include information about Axel Smith's description of the technique. As you see, he advocates only using the engine after your own analysis and annotation is complete and then using it to check your analysis, not in any way to make the analysis, try out variations, etc. Where he does advocate using the engine in the way you describe (to try out variations and explore other lines) is in his chapter on constructing an opening file as part of your opening preparation.– Brian Towers ♦Commented Jan 1, 2018 at 11:22
-
@BrianTowers That does seem to be a reasonable way of going about it (+1). The method seems to be what I thought, except it's done with a friend, not an engine. Using an engine in the first place might provide the strongest moves, but also doesn't let one put one's brain into gear, as you put it. Thanks! Commented Jan 2, 2018 at 16:18
Not a scientific answer, but several points that seem obvious to me:
Just as with many activities, in order to maintain your level in chess you need a bit of practice. Or said differently, if you stopped playing you would end up playing at lower level.
If we define "experience" as the knowledge how to deal with certain situations on the chess board, gaining new experiences through playing will get more and more difficult, because you might end up playing most of your games based on your known experiences. Also if you end up gaining a new experience it might be so specialized knowledge that you will never encounter the same situation again in a game (so cannot use it to increase ELo). So eventually playing more games will not teach you anything new, but will only help you to refresh your knowledge.
The Elo system gives the relative strength of players only. So in fact it might be possible that all players are constantly improving their absolute strength while staying at the same Elo rating.
A bracket of 200-300 Elo point is a pretty big difference in playing strength.
Playing strength depends on lots of factors such as age/fitness, opening/endgame knowledge, tactics, time invested in chess, .... Experience gained through playing might easily by canceled by any of the others.
answered Dec 30, 2017 at 12:31
-
Tank you for your input. Let me note i'm talking about active players only. The second point allow me to respectfully disagree.I thought about your third point but there are always new (inexperienced) players joining the pool so there must be some other explanation. And yes 300 points is a big difference but people ususaly go up and down and up the question is more about why we don't see positive correlation with game played. Commented Dec 30, 2017 at 13:54
It all depends on the quality of the experience. And also the frequency although quality is more important.
Playing vs the strongest engine will not help you learn and improve as much as more focused training will. You need to play someone/thing slightly better than you to improve. Playing GMs won't help you learn unless they are also giving you lessons AT THE LEVEL you can understand and use.
Perfect practice makes perfect. Making the same mistakes over and over will not help you improve. Most players get to their level and stay there until they get old.
One old chess publication had an article saying that to improve you should give up playing for a year. Focus on learning that actually helps you.
Playing is a minor part of improving. You need to focus on learning basic tactics, end games, and slightly more advanced ideas such as pawn structure, positional play, as well as openings.
Everybody hits their maximum. That may depend on talent, study, playing, amount of time available, as well as interest in getting better. So while most players improve at first, they all hit their limit. And then when they get really old they start to come back down.
200-300 is a fairly narrow range of results that accounts for the randomness of that player as well as his opponents. Some people have off days. Others get a bluebird fumblefinger and an easy win.
